2

ANALYSIS OF X-BAR THEORY

2.1 Introduction

This chapter attempts to analyze the X-bar Theol)," While this theory is claimed to be

universal making itvery suitable for the task of Machine Translation, we find a lot of

variations and inconsistencies within and between different studies in regard to the

usage of terminology, rule constraints, and the rule schemata. All this results to have

different X-bar structures for the same phrase which rather takes the X-bar Theory

away from its original goal and in any case is not desirable. The chapter raises these

issues and emphasizes the need for the standardization of the X -bar Theory. Another

related issue within the scope of the present study is the treatment of Double Object

Constructions in the light of Binary Branching Requirement. Most of the existing

studies on the topic generally skip or present very sketchy rather contradictory

treatment. The chapter proposes a solution for this problem, and provides the

justification.

Among various aspects of the X-bar Theory, the chapter covers the X-bar

structure of Double Object Constructions which most of the authors skip or give very

sketchy treatment. The chapter suggests that the distinctions between Specifier,

Adjuncts, and Complements are not only based on structural differences as suggested

in the existing literature but to the large extent are determined by the relationship of

syntax with the lexicon via Projection Principle, the Argument Structure of lexical

categories, the Theta Theory, the Case Theory and the Government Theory.

Section 2 below will focus on different X-bar levels used in X-bar Theory and

introduces the three level symmetrical rule schemata. Section 3 treats the Double

Object Constructions in the light of Binary Branching Requirement, for which, we

feel, ~e existing studies are insufficient. Section 4 concentrates on difficulties that

arise due to inconsistent use of the terms Specifiers, Adjuncts and Complements.

38

Section 5 concentrates on the different rule schemata taken in existing studies on the

topic. Section 6 provides a justification for the proposed solution in the light of Theta

Theory, Case Theory and Government Theory. The last section brings out the

conclusions and summarizes the chapter.

2.2 X-Bar Levels and the Proposed Rule Schemata

Different works in X -bar Theory have incorporated different X -bar levels. Some

works adopt the symmetrical theory of categories whereas others argue in favour of

the asymmetrical theory of categories. The symmetrical theory assumes all- categories

to permit the same range of phrasal expansions whereas the asymmetrical theory

permits different categories to have different range of phrasal expansions.

Some authors have adopted the symmetrical theory of categories. For example,

Chomsky adopts a symmetrical three-level X-bar theory (Chomsky [l981a, 1986a]).

For us, the three-levels mean XO, X', X", and not:XO, X', X", X"' as some authors take.

The three-level symmetrical X-bar analysis has been adopted, among others, in works

of Cook [1988], Duarte [1991], Freidin [1992], Haegeman [1991], and Radford

[1988]. Jackendoff in his seminal work on X-bar Syntax argues for a symmetrical

four-level X-bar Theory, namely Xo, X', X", X"' (Jackendoff [1977]). A similar

analysis is adopted in Halitsky [1975] and Binkert [1984]. Some works have used

four-bar level analysis for particular constituents. Selkirk [1977] adopts four-level

theory for analysis of Noun Phrase (NP). Akmajian, Steele, and Wasow [1979] adopts

the four-level analysis for Verb Phrase (VP). A critique of this analysis is given in

Pullum [1981], Gazdar, Pullum and Sag [t982] and Takejawa [1984]. Riemsdijk

[1978] adopts the four-level X-bar analysis for Prepositional Phrase (PP). Stuurman

[1985] has argued in favour of a symmetrical two-level X-bar Theory. Hale proposes

two-level X-bar analysis for non-configurationallanguages (Hale [1983]).

Some proponents of X-bar Syntax have argued in favour of an asymmetrical

theory of. categories. Emonds [1985] presents an asymmetrical theory in which

Maximal Projection for V is taken to be three-level but the maximal projection of

other categories is assumed to be two-level. Jacobsen [1986: 90] takes two-level

maximal projection for V, but three-level maximal projection for other categories.

39

We adopt the symmetrical X -bar analysis. This way we do not need different

sets of rules for different phrasal categories. The same set of rules will be applicable

to all phrases thus leading to uniformity and computational efficiency in terms of

efforts, time and space. Further, we need exactly three levels of projection, namely XO,

X', and X". For any head XO, we require the level X' to take care of constituents larger

than XO. Both the compulsory (Complements) and optional (Adjuncts) phrases can be

joined at X', one at a time, by Complement Rule (refer Rule (4) or Rule· (3) below) or

the Adjunct Rule (refer Rule (3) below), respectively. The highest X' which is not

dominated by any other X-projection is the maximal projection of XO denoted by XP

or X". The Specifier is attached at the X" level by the Specifier Rule (refer Rule (2)

below). Further, if there are any outer adjuncts, they could be accommodated at X"

level by Outer Adjunct Rule (refer Rule (l) below). Thus we need only three levels

namely XO, X', and X". This three-level hierarchical structure can take care of 'do so'

substitution and 'one' substitution constructs which can not be taken care of by a two­

level flat structure (Radford [1988], Haegeman [1991 D. Thus the three-level

symmetrical X-bar analysis is more suitable both linguistically as well as

computationally.

Accordingly, an X-bar phrase in a language is defined using the following rule

schemata:

*X" ~ (Y") ; X" - (1) (The Outer Adjunct Rule)

X" ~ (Z") ; X' - (2) (The Specifier Rule)

·X' ~ (T") ; X' -(3) (The Adjunct or the Complement Rule)

X' ~ (W"); XO - (4) (The Complement Rule)

The terms included in parentheses indicate that they are optional, '*' indicates that the

rule is optional, and a semicolon in the rules indicates that the terms on the right hand

side are not taken as ordered. Each of Y", Z", T", and W" are full phrases like for

example an NP or PP.

40

\

In addition, Rule (1) is referred as Outer Adjunct Rule, and will be used in

Inflectional Phrase (IP) and in Noun Phrase (NP) to handle the Outer Adjuncts, or in

phrases where Adjuncts intervene between Complements and the head (e.g. in VP and

the corresponding NP). Rule (2) is referred as Specifier Rule as taken standardly but

not in the sense, for example, of Chomsky [1981a] or Larson [1988, 1990] where they

treat one of the two internal arguments in a Double Object Construction as the

Specifier. Rule (3) is referred as the Adjunct or the Complement Rule. We do not call

this rule as Adjunct Rule as has been taken traditionally in the existing literature. It is

done so to take care of the Double Object Constructions in view of Binary Branching

Requirement and will be justified in the subsequent sections. Rule (4) is referred as

the Complement Rule. Note that there are three parameters associated with this rule

schemata namely the Specifier Parameter, the Complement Parameter, and the

Adjunct Parameter (jointly corresponding to Adjuncts in Rule (l) and Rule (3)).

In the literature, there are three conditions laid down on X-bar rule schemata

(Radford [1988: 258-78]). We consider each of them in tum.

2.2.1 The Endocentricity Constraint

The Endocentricity Constraint requires that each XP must have XO as its head. Some

authors do not follow this constraint. They argue that there are exocentric phrases in

addition to the endocentric phrases. For example, there is a difference of opinion when

dealing with Verb Phrases (VPs) which contain an auxiliary verb in addition to the

main verb. Within such a VP, while most of the studies take V as the head, but

Warner among others argues for the auxiliary to be the head and thus not respecting

the Endocentricity Constraint. He stresses, "In sentences or VPs which contain an

auxiliary, the non-auxiliary verb is not even a serious candidate for head ... " (Warner

[1993: 22]) (also see Dorr [1993b: 55-6] and Radford [1988: 282] for such a

proposal). On the contrary, Napoli states that auxiliary is not the head of VP (Napoli

[1985: 293]). Some authors, e.g. Goodall [1987] among others (see Haegeman [1991:

132], and Cook [1988: 101], for instance), treat coordinate phrases, like 'Mohan ate

Surinder' ('Mohan and Surinder'), as NP where the head is a Conjunct not Noun, thus,

again, not obeying the Endocentricity Constraint. These phrases will be treated in the

next section. Some authors (see, for example, Bhatia [1993] and Gill and Gleason

41

[1962], among others) treat phrases involving Case assigner, e.g. 'Surinder ne'

(Surinder NOM Case) or 'Mohan nuun' (Mohan in DAT or ACC Case), as an NP.

Here again the Endocentricity Constraint is not satisfied. Such a phrase must be

treated as a Case Phrase with Case assigner as the head. In every case, the

Endocentricity Constraint must be respected.

2.2.2 The Modifier Maximality Constraint

The Modifier Maximality Constraint requires that the non-heads in the expansion of a

rule must be maximal projections, i.e., each yII, Z", Til, and W" must be a full phrase

like X".

Many researchers do not follow this condition strictly, rather they question this

constraint. In addition to N, V, A, and P, we have other lexical categories, e.g.,

adverbs (ADV), conjunctions (CONJ), determiners (DET), auxiliary verbs (AU X),

quantifiers (Q), and Case assigners (CASE). We know that X projects to XP, for X =

N, V, A, P. But what about other lexical categories mentioned here. Freidin (1992: 26,

36] raises this question: "whether or not every lexical category projects its own

phrasal category is open to discussion". He takes the structure for 'the boy' as shown

in Figure 2.1.

N"

DE/~N' I I

ilie N

I boy

Figure 2.1 X-bar Structure of an NP violating the Modifier Maximal Constraint.

42

Here, N projects NP but DET does not project its own phrasal category, rather DET is

a lexical constituent of the phrasal projection NP of N. Pinker also treats determiners

as minor lexical categories, with no superordinate projections (Pinker [1984]). There

are numerous instances in Haegeman where Specifier (determiner) is not a full phrase

(e.g. see Haegeman [1991: 87-95]). In fact Haegeman does not impose this condition

on Specifier in Rule (2). Cook also does not treat 'the' (DEn as a full phrase;

Specifier does not always consist ora complete phrase in its own right (Cook [1988:

99]). Thus, the Modifier Maximality Constraint is not satisfied for DET in these

works. Some authors while giving structure for VP take Z" in Rule (2) as aspectual

auxiliary, e.g. in Radford [1988: 231, 237]. Here, too, the Modifier Maximality

Constraint is violated. Similar argument holds for other lexical constituents, e.g., Q,

and AUX denoting modal, aspect and tense. For these categories, the issue is not so

clear. Thus it is an open question whether every lexical category projects its own

phrasal category. Because of this we see that X-bar analysis has generally been

applied to NP, VP, AP, and PP only; A few authors have extended the concept to

other phrases, namely Adverb Phrase, Case Phrase, and Conjunct phrase, etc. (see

Radford [1988] for Adverb Phrase, Berwick and Fong [1990: 302] for Case Phrase,

Larson [1990: 596] and the references sighted therein for Conjunct Phrase).

Thus, we either need to treat these lexical categories as full phases. For .

example we can take AUX as AUXP (Auxiliary Phrase) with structure as shown in

Figure 2.2.

43

AUX"

AUX'

AUX

IS

Figure 2.2 A possible structure for Auxiliary Phrase.

Alternately, we could relax the Modifier Maximality Constraint. For uniformity, we

decide to follow the former view. The rule schemata allows this.

2.2.3 The Category Neutrality Constraint

The Category Neutrality Constraint requires that the rule. schemata must be

formulated entirely in terms of category variables. This means that the Rules (1-4) are

applicable for all X where X can be any head. This constraint automatically follows in

our case.

The proposed rule schemata satisfies all these three constraints.

2.3 Binary Branching Requirement and Double Object Constructs

Binary Branching Requirement (or Hypothesis) is an important requirement on

syntactic configurations. The concept was originally incorporated within. the

framework of Government and Binding Theory by Kayne (Kayne [1983]). Kayne

proposes that binary branching constitutes an upper limit on syntactic configurations,

and the maximal number of children a node can take is therefore two. This

44

requirement guarantees that paths relating a trace or an anaphora to its antecedent are

not ambiguous. Further, a grammar following binary branching hypothesis is more

constrained and requires fewer decisions and leads to speedy construction of the

grammar than in a grammar with n-ary branching. It is especially important for

language acquisition.

Many authors, e.g. Aoun and Li [1989], Chomsky [1981a], Duarte [1991],

Freidin [1992], Haegeman [1991], and Larson [1988, 1990] incorporate Binary

Branching Requirement in their works, whereas others, e.g. Cook [1988], Dorr

[1993b: 136, 143], Goodall [1987], Iwakura [1987], Kajiwara [1989], Oehrle [1976],

and Radford [1988], do not take this issue into consideration.

Though some works intend to impose the Binary Branching Requirement on

syntactic configurations, it can not be satisfied there in the strict sense. It is due to the

reason that they impose the 'sisterhood condition' on Complements: "Complements

are sisters of the head". Obviously, this is contradictory. This is what happens, for

example, in Duarte (see Duarte [1991: 2&, 33]) and Haegeman (see Haegeman [1991:

95]). Note that we do not impose the 'sisterhood condition' on Complements.

Now, let us consider Double Object Constructions. Double Object

. Constructions are 3-place predicates, or equivalently· they have two internal

arguments. They are also referred to as the ditransitive lexical categories. Double

Object Constructs have been well discussed, for example in the works of Barss and

Lasnik [1986], Chomsky [1981a], Iwakura [1987], Jackendoff [1990b], Kajiwara

[1989], Larson [1988, 1990], and Oehrle [1976]. None of the studies on the topic is

sufficient. According to Duarte, a syntactic configuration is invalid if it does not

follow the Binary Branching Requirement. The solutions by Iwakura [1987], Kajiwara

[1989], and Oehrle [1976] do not meet the Binary Branching Requirement, the

solutions by others have been well criticized. The rule schemata in existing studies on

X-bar Theory for example Cook [1988], Haegeman [1991], Radford [1988], etc. are

simply incapable of handling the Double Object Constructions because it can not

incorporate the Binary Branching Requirement. The root of the probiem is the

45

imposition of 'sisterhood condition' on Complements due to Hornstein and Lightfoot

[1981 ].

Consider a Double Object Construction XO. Let its two Complements be Til

and W". If we impose the Binary Branching Requirement becaUse of its obvious

advantages, then only one of the Complements, say W", of XO can be accommodated

at the Complement position by Rule (4). We are forced for the upward branching of

X-bar structure to secure an argument position for the second Complement T" at the

so called Adjunct node. This node now needs proper interpretation. We do not call

this node as the Adjunct node, rather, first, we consult the argument structure (the

lexicon) for the entry XO. IfXO is ditransitive, the so called Adjunct node is to be taken

as the second Complement position. This is the reason we do not call Rule 3 as the

Adjunct Rule, .rather we refer it as the Complement or the Adjunct Rule. In case XO is

ditransitive, Rule J will act as Complement Rule. In case the Adjuncts are actually

present in the phrase XP under consideration, this Rule 3 will, in addition, act.as the

Adjunct Rule.

Note that the above procedure (and hence the proposed rule schemata) is

capable of handling I-place and 2-place predicates in addition to 3-place predicates.

46

In the light of Binary Branching Requirement, an X-bar structure for a Double

Object Construction yo like 'lend', 'give', 'put' etc. as taken, for example, by Cook

[1988], Dorr [1993b], and Radford [1988] as shown in Figure 2.3 will be represented

as in Figure 2.4.

V"

V'

~ yo T" W"

Figure 2.3 An invalid X-bar phrase structure for a Double Object Construct not obeying Binary Branching Requirement.

V"

I A A T"

yo . W"

Figure 2.4 X-bar phrase structure of a Double Object Construct in the light of Binary Branching Requirement.

Note that the structures like in Figure 2.3 are ruled out due to Duarte in the light of

Binary Branching Requirement.

47

An important issue related with Binary Branching Hypothesis is the structure

for Coordinate Phrases such as 'the boy and the girl'. Goodall [1987] n:eats such

phrases as NP with ternary branching structure as shown in Figure 2.5.

Figure 2.5 The structure of a Coordinate Phrase due to Goodall [1987], not satisfying Binary Branching Requirement.

Such a phrase should not be treated as NP rather it should be treated as Conjunct

Phrase with Conjunct as the head, and having the structure as shown in Figure 2.6.

Figure 2.6 A possible X-bar phrase structure for the Conjunct . Phrase, see Larson [1990: 596].

48

2.4 Specifier, Complement, and Adjunct -

We find many inconsistencies within and between different authors regarding the

definitions of Specifier, Adjunct, and Complement. Most of the works use these terms

very arbitrarily and indiscriminately.

Freidin [1992: 40] does not take Specifier into consideration in his general

schema for phrase structure. But, on page 27 (and also at pages 41 and 321), he treats

DET 'the' as Specifier: "lexical constituents of phrasal projections like Det and Adv

act as modifiers of the lexical head of the projection. Each can be referred to as a

Specifier". However, on page 40, he treats Det as Adjunct: "Adjuncts are generally

phrases (e.g. AP, PP, S) or .lexical constituents (e.g. Det for NP)". Authors such as

Cook, Haegeman, and Radford treat DET 'the' as Specifier. Similarly, Freidin treats

'very' in 'very quickly' as modifier, but other works, e.g. Radford [1988], treat it as a

Specifier. This indiscriminate or inconsistent usage of the terminology is not desirable

as this leads to different syntactic structures for the same phrase.

Consider the D-structure and S-structure configurations for VP in each of the

sentences.(a) 'John sent a letter to Mary' and (b) 'John sent Mary a letter' as taken by

Larson (see Larson [1988: 342-43, 353] and Jackendoff [1990b: 436-37]), and

reproduced in Figure 2.7, Figure 2.8 and Figure 2.9, Figure 2.10 respectively. In

Figure 2.7 and Figure 2.8, the direct object 'a letter' is treated as Specifier, :whereas in

Figure 2.9 and Figure 2.10 it is treated as Adjunct. In all cases it has to be a

Complement.

49

VP

/ Spec V' /

v VP

I / L~ v/

1 Lary Figure 2.7 D-structure for VP in 'John sent a letter to Mary' as taken by

Larson [1988].

VP

/ Spec V'

V' 1

I "'""----send +--t

Figure 2.8 S-structure for VP in 'John sent a letter to Mary' as taken by Larson [1988].

so

VP

/ . Spec V'

e e

Figure 2.9 D-structure for VP in 'John sent Mary a letter' as taken by . Larson [1988].

VP

/ Spec V' /y,

V' 1

I send

V'

Figure 2.10 S-structure for VP in 'John sent Mary a letter' as taken by Larson [1988].

51

While dealing with modifiers, Larson [1988: 350] (refer pages 345-46, foot

note 11; pages 349-50, and 382; page 384, foot note 49) concludes that time and

manner PPs are not "outermost adjuncts (as is standardly assumed) but rather must be

complements". He gives following structure (Figure 2.11) for 'writes a letter to Mary .

VP

/'" SPEC V' V'

/", V

e V'

~n~ L~o~i~~

Figure 2.11 A structure for 'write a letter to Mary in the morning', where a Adjunct PP 'in the morning' is treated as Complement of V Larson [1988].

in the morning' (see Larson's structure (ii), foot note 11, pages 345-46). We note that

here the time modifier is treated as a Complement rather than as an Adjunct.

As another example, let us see the X-bar structure for Double Object

Construction due to Chomsky [1981a] (see also Barss and Lasnik [198"6], Duarte

52

VP

V,ANP2 (~l~ te lary

Figure 2.12 X-bar structure for 'gave Mary a letter' due to Chomsky [1981a].

C-MAX

A ADV C-MAX

I I " Yesterday C I-MAX

I / I "-eN-MAX I V-MAX

I I /"-John e V N-MAX

I / "-saw N-MAX P-MAX

N-J L J ~-MAX Mary,t ~ ~ it N-~ ~

I I I I older sIster lier car

Figure 2.13 X-bar phrase structure for 'Yesterday John saw Mary's older sister in her car', Dorr [1993b: 52].

[1991], Haegeman [1991: 132], lackendoff [1990b], Larson [1988: 337], and Larson

[1990: 590]), as shown in Figure 2.12.

53

Here, the second internal argument of V, the direct object, is treated as Specifier

instead of as a Complement.

Interestingly, X-bar phrase ~cture representations for the same sentence

'Yesterday John saw Mary's older sister in her car.'

are not identical in DOff [1993b: 52] and in DOff [1991b: 3] (see Figure 2.13 and

Figure 2.14). In Figure 2.13, 'Yesterday' is taken at C-MAX level whereas in Figure

2.14, it is taken as a part ofI-MAX.

/l-MAX

e I-MAX

/ " ADV I-MAX

1/1 Yesterday N-MAX e

I John

I I olaer slster

I her car

Figure 2.14 X-bar phrase structure for 'Yesterday John saw Mary's older sister in her car' , from DOff [1991 b: 3].

54

Some authors place Adjuncts generally taken at X' level to XP level. For

example Pinker [1984] takes the following structure (see Figure 2.15) for VP in '10hn

hit Mary in the park' as follows:

V" /" V' P" A I

V N" P' I I /" hit N' P N"

I I A N III Det N'

I I I Mary the N

I park

Figure 2.15 The Adjunct 'in the park' taken at V" level instead of V' level, Pinker [1984].

Larson [1990: 591] while giving the X-bar representation for '10hn visited few

friends any day this week.', also, puts the adverb 'any day this week' at VP level (see

Figure 2.16).

55

IP /" NP I' I _/"

John I VP

/ " VP NP

/" /any~ V NP ~ay this we~ I /few"

visit L friends ~ .

Figure 2.16 The Adjunct 'any day this week' taken at V" level, not at V' level, Larson [1990].

Generally, these Adjuncts are taken at V' level (see, for example, Haegeman [1991],

Radford [1988]).

Barriers assumes the existence of zero or more Specifiers whereas other works

do not take more than one Specifier.

These different treatments of Specifier, Adjuncts, and Complements lead us to

have different X.,bar structures for same construction in different works. The problems

sighted above as regards to Specifier, Adjunct, and Complement are partially due to

the definitions of the Specifier, Adjunct and Complement, and the associated rule

schemata. Larson treats one of the Complements as Specifier and uses the Specifier

Rule (2) instead of the Complement Rule (3), leading to the different structure. In case

of Double Object Construction, we take both the Complements as internal. arguments

(also taken, for example, in Duarte 11991], Grimshaw [1990], and Haegeman [1991 ]),

whereas Chomsky [1981a] and Larson [1988, 1990] take one of the arguments as

Specifier (external argument). This attributes to different X-bar structures for the same

phrase.

56

For uniformity, we need to define precisely each of the Specifier, Adjunct, and

Complements.

Let us define the terms Specifier, Complement, and Adjunct. A Complement

is an internal argument of the head. It is both c-selected and s-selected. A complement

is always a compulsory construct. A Specifier is s-selected but not c-selected. It is an

external argument of the head. An Adjunct is neither c-selected nor s-selected. An

Adjunct is an optional construct. An Adjunct is a non-argument construct (i.e. neither

external nor internal argument). There are three parameters associated with X-bar

Theory: the Specifier Parameter, the Adjunct Parameter, and the Complement

Parameter. For each phrase in a language, Specifier, Adjunct and Complement are to

be defined more precisely in terms of their parametric values.

Hornstein and Lightfoot [1981: 21] defines Specifier, Adjuncts and

Complements in terms of structural differences (also see Duarte [1991: 28],

Haegeman [1991: 95], and Radford (l988: 176, 187, 193, 196-97,279]):

(a) Specifiers are sisters of X' and daughters of X".

(b) Adjuncts are both sisters and daughters of X'.

(c) Complements are sisters of x:o and daughters of X'.

This definition for Complements violates the Binary Branching Requirement for

Double Object Constructions. But for this, we have proposed a solution above.

2.5 Different Rule Schemata

We observe different authors assume different X-bar rule schemata. They use one or

the other rule without mentioning them explicitly. Larson (1990: 591] represents 'John

visited few friends any day this week." as shown in Figure 2.1? This structure uses a

rule of the form XP -7 XP ZP (with X = V and Z = N), but the author does not include

this rule in the rule schemata taken by him. Also the rule X' -7 X' TP, the so called

Adjunct Rule, is not taken at all.

57

Barriers takes the rule schemata as:

X"=X"*X'

X'=XX"*

Note that the Adjunct Rules are absent here, however the book mentions about

Adjuncts repeatedly.

Radford [1988: 277] takes the rule schemata as follows:

X" ~ X', (YP)

X' ~ X', (YP)

X' ~ X, (YP)*

(Specifier Rule)

(Adjunct Rule)

-(A)

-(B)

(Complement Rule) -(C)

Radford does not include Rule (I) taken here (see Section 2). This rule schemata

cannot satisfy Binary Branching Hypothesis for Double Object Constructs. We treat

Rule (B) here as Adjunct or Complement Rule depending on whether X is non­

ditransitive or ditransitive. Here it is treated as Adjunct Rule.

Haegeman [1991: 95, 369-70] takes the rules as:

Xp* ~ XP; YP

XP ~ Spec; X'

X'· ~ X'; yp

X' ~X;YP

Note that second rule here indicates that Spec is not treated as full phrase. Haegeman

[1991: 95] mentions, that 'Complements combine with X to form X' projections'. The

author overlooks the case of Double Object Constructions with Binary Branching

Hypothesis in mind. This statement indicates: (a) Complement is defmed structurally

(b) Binary Branching Requirement can not be satisfied, (c) This rule s~hemata is

different from the one proposed by us though this looks to be the same. This rule

58

schemata differs our one in tenns of usage. We treat Rule (3) here as Adjunct or

Complement Rule depending on whether X is non-ditransitive or ditransitive. Here it

is treated as Adjunct Rule.

Cook [1988: 103] takes the Rule Schemata as:

X" -7 specifier X'

X' -7 X complements

Note here that there are no Adjunct Rules. Further this rule schemata cannot

satisfy Binary Branching Hypothesis for Double Object Constructs.

Thus different authors take some or the combination of the four rules thus

leading to different X-bar structures for the same phrase. Further, these rules do not

indicate where the second complement in a Double Object Construction should be

attached.

Since the rule schemata is the basis for assigning the syntactic phrases, these

different rule schemata adopted by different authors obviously will lead to dissimilar

X-bar phrase structure for the same sentence. If X-bar Theory is to be universal and to

be applicable cross-linguistically, the rule schemata must be made standard.

2.6 Justification of the Solution Proposed

Keeping the different issues in mind, we have proposed to adopt three-level

symmetric X-bar rule schemata. These rules are rewritten below:

·X" -7 (Y") ; X" - (I) (The Outer Adjunct Rule)

X" -7 (Zit) ; X' -(2) (The Specifier Rule)

·X' -7 (Tit) ; X' - (3) (The Adjunct or the Complement Rule)

X' -7 (Wit); x<> -(4) (The Complement Rule)

The associated X-bar structure is shown in Figure 2.17.

59

X"

/~ Y" X"

/~ -Z" X'

/~-X' T"

~/~T" Figure 2.17 A well-fonned syntactic structure confonning to

the proposed rule schemata.

The rule schemata and the proposed X-bar structure is capable of handling not only up

to 3-place predicates (as found, for example, in English) but is also capable of

handling 4-place or in general n-place predicates (In contrast to English, Punjabi and

Hindi, as will be seen in next chapter, have 4-place predicates also which correspond

to causative fonns of the Verb). Note that in the rule schemata and the corresponding

X-bar structure here, T" is not necessarily the Adjunct rather it is decide~ by the

Argument structure of the lexical item Xo.

Applying the proposed Rule Schemata, the canonical or the basic -X-bar

structure (i.e. leaving aside the Adjuncts) for a I-place, 2-place and 3-place. predicates

(Double Object Constructs) respectively look as shown in Figure 2.18, Figure 2.19,

and Figure 2.20 below.

60

X"

Spec X'

Figure 2.18 Canonical X-bar structure of a I-place predicate.

X"

/"" Spec X'

/"'" XO W"

Figure 2.19 Canonical X-bar structure of a 2-place predicate.

Figure 2.20 Canonical X-bar structure of a 3-place predicate.

61

Though we have deviated from the structural differences between a

Complement and an Adjunct due to Hornstein and Lightfoot [1981] but this departure

is justified as argued below. Note that we do not impose the 'sisterhood condition' on

the Complements.

"The phrases in a sentence are tied into the lexicon via their heads" (Cook

[1988: 95]). There is a close interaction between syntax, Theta Theory, Case Theory,

the concept of Government, the argument structure, and the Projection Principle.

Chomsky mentions: "I omit consideration here of possible further structure and

assume that the basic properties of phrase for particular languages are determined by

fixing parameters of case theory and theta-theory and by lexical properties" (Chomsky

[1986a: 3]). This implies X-bar structure is not only determined by the X-bar rules but

must be validated in terms of Theta Theory, Case Theory, and the concept of

Government.

Obviously, in case of one-place or two place predicates we do not have any

problem. We need to justify our solution for three or n-place predicates.

We attempt to justify below the proposed solution in the light of each of Theta

Theory, Case Theory and the Government Theory.

2.6.1 Theta Theory

There is a relationship between syntax and theta-roles. Syntax and theta-roles must be

integrated. The structure proposed for the 3-place predicates (Double Object Constructs)

satisfies the Theta Criterion otherwise the Projection Principle will get violated, i.e., the

arguments will not get saturated. The Theta Criterion requires: "Each argument bears

one and only one a-role, and each a-role is assigned to one and only one argument"

(Chomsky [1981 a: 36]). The Projection Principle requires that the thematic (argument)

structure associated with lexical items must be saturated (realized) in the syntax:

"Representations at each syntactic level (i.e., LF, and D- and S-structure) are projected

from the lexicon, in that they observe the subcategorization properties of lexical items."

(Chomsky {1981a: 29]), or equivalently "Lexical structure must be represented

62

categorially at every syntactic level" (Chomsky [1986b: 84]). In the proposed structure,

the head:XO assigns internal theta-roles to both (each in case ofn-place predicates) of its

internal arguments and the external theta-role to its external argument. It satisfies (a)

Duarte's statement: "No XO can assign an internal 8-role to a position outside its one

bar-level projection" (Duarte [1991: 38-39]), and (b) Cook's statement: "The Agent is

an external 8-role that goes outside the maximal projection of the verb; other roles

such as Goal, or Patient are internal within the maximal projection" (Cook [1988:

115]). The condition (a) here does not stop us to assign internal theta-role to the

second internal argument, and the condition (b) here confirms this. In this light our

proposed solution (the rule schemata as well as the structure) is justified. Moreover,

both the Complements are governed by :xo (as will be seen below) implying thereby.

:xo theta-governs (i.e. it assigns internal theta-role and governs) both (each) of its

Complements. We cannot assign the second Complement the external 8-role due to

the fact that "Only the external 8-role is not governed by the verb" (Cook [1988:

154]), or in general by any head. In our case both the 8-roles are governed by XO, and

therefore must be internal. The argument is further strengthened by the fact that "8-

roles are assigned under government" (Duarte [1991: 41 l, and Riemsdijk and Williams

[1986: 242]). Note that we differ here from Chomsky where he assigns external 8-role

to one of the internal arguments. It happens so because he treats the second

Complement as the Specifier (the external argument), and accordingly he assigns

different structure.

For example, in a sentence like 'Mohan Surinder nuun kitaab de vegaa'

(,Mohan will give Surinder a book'), the entry: de[-, CASEP, NPl <Agent, Goal, .

Patient> for the ditransitive verb 'de' ('give') specifies three theta-roles: Agent, Goal,

and Patient (also sometimes referred to as Theme). The internal arguments CASEP

'Surinder nuun' and NP 'kitaab' are both directly theta marked and are respectively

assigned the internal theta-role of Goal, and the internal theta-role of Patient. NP

'Mohan', the external argument, is assigned the external theta-role of Agent to

Grammatical Function (GF) subject. This also satisfies the argument hierarchy due to

Grimshaw (see Grimshaw [1990]). The head 'de' theta-governs both of its internal

arguments (Complements) namely CASEP 'Surinder nuun' and NP 'kitaab'. The

external argument NP 'Mohan' is not governed by 'de' and hence not theta-governed.

63

2.6.2 Case Theory

There are two requirements of Case Theory: (a) Case Filter, and (b) Adjacency

Requirement on Case Assignment. Let us see that each of these criterion is satisfied

by our proposed solution.

The Case Filter requires that each NP must be assigned a Case. Let us

visualize the possible appearances (positions) of an NP in a sentence:

(i) it can be part of an NP (like in 'Mohan's brother). In this case, the Case

assigner" 's " assigns Genitive Case to the NP. We treat 'Mohan's' as Case

Phrase not as NP like for example in Dorr [1993 b: 52] or Haegeman [1991: 119].

We do not treat 'Mohan's' as NP as in this case there will be no one to assign the

Case to this NP.

(ii) it can be a constituent of a Prepositional/ Postpositional Phrase or Case Phrase. In

this case the Preposition! Postposition or the Case Assigner in the respective

phrase will assign respectively the oblique or ACCUSATIVE Case to the NP.

Note that we treat phrases with daa, ne, and nuun as postpositioris as Case

Assigners since the respective Case Phrase leads to give a grammatical function

like Agent or Goal, but not as postpositions as the respective Postpositional

Phrase acts as an Adjunct.

(iii) it can be part of AP (like in 'beautiful than Preet'). Here again preposition!

postposition will assign the Case to NP.

(iv) it can be a part ofVP, where V will assign the ACCUSATIVE Case to the NP.

(v) NP could be the subject Grammatical Function (GF). The Subject GF is assigned

the NOMINATIVE Case by I.

(vi) Now think of a Double Object Construction where both Complements are NPs

(like 'He sent Mary a letter'). V assigns ACCUSATIVE Case to one of the

64

Complement NP. The other Complement NP is treated as Case Phrase with Case

Assigner as Null (see Iwakura [1987], Kajiwara [1989]).

Thus in all cases, each NP is assigned the Case, hence satisfying the Case Filter.

The Adjacency Requirement on Case Assignment expects that the Case

assigner must not be separated from the Case assignee which they case-mark by the

intervening Adjuncts. If the internal arguments of a head XO uniformly originate in the

order specified in the Argument Structure for XO, then XO will assign the structural

ACCUSATIVE Case to its direct object, and the Case assigner (or the Preposition in

case of English) (which may be Null as mentioned above) will assign the Case to the

indirect object of XO. In case the constituents in a phrase do not follow the Adjacency

Requirement, this is considered to be a derived order. In such a case, internal

arguments are taken to the left! right of Adjunct(s) determined by the parameter

settings of the head with a trace at the Complement positions and are adjoined to the .

XP (Haegeman [1991: 540,543]). In such a situation the internal arguments are case­

marked via their traces. Note that the Cases are assigned under the structural

relationship of government.

As an example, in a sentence like

'Mohan Surinder nuun kitaab devegaa. '

Mohan Surinder DA T book give-fut.ms

(,Mohan will give Surinder a book. ')

with 'Surinder nuun kitaab devegaa' as VP the Double Object Construct 'deNaa'

('give') assigns ACCUSATIVE (Objective) Case to the NP 'kitaab', and the Case

assigner 'nuun' assigns Case to the NP 'Surinder'.

In a sentence like

'Mohan Surinder nuun kitaab kall devegaa.'

Mohan Surinder DAT book yesterday give-fut.JV.s

(,Mohan will give Surinder a book tomorrow. ')

65

where the Adjacency Requirement on Case Assignment is not met, the structure will

be

'Mohan Surinder nuun kitaabi kall t· 1 devegaa.'

Note that some authors treat a phrase like 'Surinder nuun' ('to Surinder') as

Postpositional Phrase (PP), others as Noun Phrase (NP) (According to Bhatia, and

Gill, ne, nuun, and daa mark NPs).1t is wrong to treat it as PP as PPs act as Adjuncts.

It is wrong to treat this as NP, because if treated so there will be no one to assign Case

to this NP. We treat this phrase as Case Phrase. The markers ne, nuun, and daa act as

grammatical functions and therefore must be treated distinctly from other

prepositions! postpositions. Phrases with these prepositions! postpositions should be

treated as Case Phrases (JackendofT[1977: 80-81]).

2.6.3 Government Theory

The concept of Government also influences the structure of a phrase: "The projections

from the lexical entry onto the syntax also depend on government." (Cook [1988:

153]). "The verb governs the elements that it projects onto the sentence" (Cook [1988:

35]). This statement is true for all lexical heads; a head must govern its

Complement(s ).

In a Double Object Construction, the head XO governs each of its internal

arguments (Complements). XO will not govern the internal constituents of its

Complements. They will be governed by their respective heads. XO will not be able to

govern its external argument (the Agent) as it can not m-command the external

argument: there will be a maximal projection XP of X which dominates X and does

not dominate the external argument.

In the sentence 'Mohan ne Surinder nuun kitaab diltil" with 'Surinder nuun

kitaab dittil"' as VP, V 'de' governs both of its internal arguments (direct 9bject NP

'kitaab', and indirect object the Case Phrase 'Surinder nuun') since (i) it is a governor,

(ii) it m-commands them, and (iii) the minimality condition is satisfied. V does not

govern any element within 'Surinder nuun' since the barrier Case Phrase intervenes.

66

Within 'Surinder nuun', 'nuun' will govern 'Surinder'. The head V ('de') will not be

able to govern its external argument 'Mohan ne' as the external argument is not m­

commanded by V: there is a m(;lXimal projection VP of V which dominates V and

does not dominate the external argument 'Mohan ne'. In other words·VP will act as a

barrier. Thus, the structure generated by X-bar rules satisfies the definition of

Government (Haegeman [1991: 152], and Duarte [1991]).

2.7 Conclusions

In trying to provide an answer to the raised issues the chapter has concentrated more

on illustrating the types of difficulties that the X-bar Theory has due to the highly

indiscriminate and ad hoc use of the terminology and the rule schemata which leads to

different syntactic configurations for the same phrase. The aim of the present study is

not to criticize the X-bar Theory or to conclude that this theory be abandoned, but it

realizes the need for the standardization of the X-bar Theory and sets forth certain

issues for the linguistic community so that the X-bar Theory is consistent and

applicable cross-linguistically in the true sense.

We have proposed to adopt the three-level symmetric X-bar Theory for its

linguistic and computational motivations.

Another important issue within the X-bar Theory is the treatment of Double

Object Constructions. The existing studies work well for I-place or 2-place

predicates, they are simply insufficient for 3-pace predicates (the Double Object

Constructions). A major problem in dealing with such constructions has been that the

existing studies take Complements to be sisters of the head. This obviously cannot be

satisfied in view of Binary Branching Requirement. We propose a solution for the

Double Object Constructions which brings us out of this dilemma. We have deviated

from the existing standard general notion of structural differences between Adjuncts

and Complements, but this departure is not as much as found elsewhere in GB

Theory. The chapter concludes that this distinction not merely be b~ed on the

structural differences but on more fundamental principles of GB Theory. In fact these

differences remain no longer relevant when we deal with Double Object Constructions

in the light of Binary Branching Requirement.

67